Frequent Long-Range Epigenetic Silencing of Protocadherin Clusters on 5q31 in Wilms’ Tumor

Anthony R. Dallosso1, Anne L. Hancock1, Marianna Szemes1, Kim Moorwood2, Laxmi Chilukamarri1, Hsin-Hao Tsai1, Abby Sarkar3, Jonathan Barasch3, Raisa Vuononvirta4, Chris Jones4, Kathy Pritchard- Jones4, Brigitte Royer-Pokora5, Sean Bong Lee6, Ceris Owen1, Sally Malik1, Yi Feng7, Marcus Frank8, Andrew Ward2, Keith W. Brown1*, Karim Malik1* 1 Cancer and Leukaemia in Childhood – Sargent Research Unit, Department of Cellular and Molecular Medicine, School of Medical Sciences, University of Bristol, Bristol, United Kingdom, 2 Department of Biology and Biochemistry, University of Bath, Bath, United Kingdom, 3 Department of Medicine, Columbia University College of Physicians and Surgeons, New York, New York, United States of America, 4 Paediatric Oncology, Institute of Cancer Research/Royal Marsden NHS Trust, Sutton, United Kingdom, 5 Institute of Human Genetics and Anthropology, Heinrich-Heine University of Duesseldorf, Duesseldorf, Germany, 6 Genetics of Development and Disease Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, Maryland, United States of America, 7 Department of Biochemistry, School of Medical Sciences, University of Bristol, Bristol, United Kingdom, 8 Developmental Biology, Institute of Biology 1, University of Freiburg, Freiburg, Germany

Abstract Wilms’ tumour (WT) is a pediatric tumor of the kidney that arises via failure of the fetal developmental program. The absence of identifiable mutations in the majority of WTs suggests the frequent involvement of epigenetic aberrations in WT. We therefore conducted a genome-wide analysis of promoter hypermethylation in WTs and identified hypermethylation at chromosome 5q31 spanning 800 kilobases (kb) and more than 50 . The methylated genes all belong to a-, b-, and c- protocadherin (PCDH) gene clusters ( Organization nomenclature PCDHA@, PCDHB@, and PCDHG@, respectively). This demonstrates that long-range epigenetic silencing (LRES) occurs in developmental tumors as well as in adult tumors. Bisulfite polymerase chain reaction analysis showed that PCDH hypermethylation is a frequent event found in all Wilms’ tumor subtypes. Hypermethylation is concordant with reduced PCDH expression in tumors. WT precursor lesions showed no PCDH hypermethylation, suggesting that de novo PCDH hypermethylation occurs during malignant progression. Discrete boundaries of the PCDH domain are delimited by abrupt changes in histone modifications; unmethylated genes flanking the LRES are associated with permissive marks which are absent from methylated genes within the domain. Silenced genes are marked with non-permissive histone 3 lysine 9 dimethylation. Expression analysis of embryonic murine kidney and differentiating rat metanephric mesenchymal cells demonstrates that Pcdh expression is developmentally regulated and that Pcdhg@ genes are expressed in blastemal cells. Importantly, we show that PCDHs negatively regulate canonical Wnt signalling, as short-interfering RNA–induced reduction of PCDHG@ encoded leads to elevated b- catenin , increased b-catenin/T-cell factor (TCF) reporter activity, and induction of Wnt target genes. Conversely, over-expression of PCDHs suppresses b-catenin/TCF-reporter activity and also inhibits colony formation and growth of cancer cells in soft agar. Thus PCDHs are candidate tumor suppressors that modulate regulatory pathways critical in development and disease, such as canonical Wnt signaling.

Citation: Dallosso AR, Hancock AL, Szemes M, Moorwood K, Chilukamarri L, et al. (2009) Frequent Long-Range Epigenetic Silencing of Protocadherin Gene Clusters on Chromosome 5q31 in Wilms’ Tumor. PLoS Genet 5(11): e1000745. doi:10.1371/journal.pgen.1000745 Editor: Veronica van Heyningen, Medical Research Council Human Genetics Unit, United Kingdom Received March 30, 2009; Accepted October 29, 2009; Published November 26, 2009 This is an open-access article distributed under the terms of the Creative Commons Public Domain declaration which stipulates that, once placed in the public domain, this work may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. Funding: This work was supported by Cancer and Leukaemia in Childhood - Sargent, the Children’s Leukaemia Trust, and Kidney Research UK. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: [email protected] (KWB); [email protected] (KM)

Introduction Several epigenetic lesions have previously been identified in Wilms’ tumour, in particular loss of imprinting at chromosome Wilms’ tumour (WT) represents a paradigm for cancer arising 11p13 [2,3] and 11p15 [4], which we have shown to be early and from disrupted development. Failure of the metanephric blaste- independent events [5]. In common with other cancers, WTs also mal cells to undergo mesenchymal to epithelial transition, show tumour suppressor gene silencing which includes genes such together with proliferation of these undifferentiated cells, is as HACE1 (73% of tumours analysed) [6], RASSF1 (56%) [7], intrinsic to the development of Wilms’ tumours [1]. Thus WT CASP8 (43%), MGMT (30%), RASSF5/NORE1 (15%), and predisposing genes are often critical in normal nephrogenesis. As CDKN2A (10–15%) [8]. In addition, we have recently shown the aetiology of WTs cannot be explained solely by the known over-expression of the GLIPR1 gene resulting from promoter genetic changes, we have evaluated epigenetic changes in WTs. hypomethylation (87%) [9].

PLoS Genetics | www.plosgenetics.org 1 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

Author Summary tumour had between 12.0% and 33.7% (mean 23.8%) of hypermethylated targets in common with the other tumours The development of tissues and organs in the human body analysed. The genomic distribution of recurrently hypermethy- requires carefully regulated production of proteins by cells. lated genes was non-random (P,0.001, chi square test), with many Proteins permit the growth and development of the many being located together or within gene clusters (Table S1). varied structures required for a healthy body. In many Contiguous hypermethylated genes were more likely to be found diseases, including some cancers, tissues and organs fail to in several tumours (13 out of 17 loci). develop as they should due to the normal production of Strikingly, of 146 annotated genes hypermethylated in 3 or proteins being changed. The work presented here shows more tumours, 25 were PCDHs located at chromosome 5q31 and that in Wilms’ tumor, a childhood cancer of the kidney, a MeDIP-chip data revealed enriched tumour/normal ratios large group of related proteins that are likely necessary for growth and development of a normal kidney are not across 50 closely-linked CGIs spanning 800 kb on chromosome produced properly. This is due to their production being 5q31.3 (Figure 1A and Figure S1C). Hypermethylation at this switched off within the cancer cells. We show how these locus was identified in all tumours examined (Table S2) and proteins, known as protocadherins, can themselves alter hypermethylated CGIs localise to PCDHA@, PCDHB@ and the function of other proteins already known to be [email protected]@ gene is encoded by a single, unspliced important in normal growth and cancer. Thus our study exon, while the PCDHA@ and PCDHG@ genes have unique 59 increases our understanding of how protocadherins are exons splicing into cluster-specific constant region exons important in normal growth and of how altering proto- encoding invariant cytoplasmic domains [10]. Proximal and cadherins may lead to disease, such as cancer. distal boundaries to the hypermethylated domain were apparent (Figure 1A) and, remarkably, tumour MeDIP signal was enriched at virtually all of the PCDH CGIs across the locus (Figure 1B). As In order to identify candidate genes involved in Wilms’ a complementary screening approach, we also conducted tumorigenesis, we undertook genome-wide analysis of promoter enrichment of methylated DNA from the WiT49 Wilms’ tumour methylation. We have identified pronounced tumour-specific cell-line [18] using recombinant methyl-CpG binding domain hypermethylation in a region spanning ,800 kb of chromosome (MBD) protein, which permitted microarray methylation profil- 5q31. This region contains members of the PCDH superfamily in 3 ing without probe DNA amplification. This analysis also showed multi-gene clusters (PCDHA@, PCDHB@ and PCDHG@) [10]. hypermethylation of multiple PCDHs across the region (data not Hypermethylation of this locus is an example of long range shown). Both genome-wide analyses identified genes previously epigenetic silencing, which has previously been reported in known to be hypermethylated in Wilms’ tumour, such as RASSF1 colorectal cancer at chromosome 2q14.2 [11], the MLH1 locus (Table S2). Thus our genome-wide promoter analysis identifies a on 3p22 [12], and for the HOXA gene cluster on chromosome novel region of long-range epigenetic silencing in WTs on 7p15 in breast and lung cancers [13,14]. We demonstrate that chromosome 5q31.3. silencing of is concomitant with DNA hyper- methylation and repressive histone modifications. Although little is Long-range epigenetic modifications at 5q31 are known about the functions of these clustered PCDHs, other members of the PCDH superfamily have been shown to have common in Wilms’ tumour but not in preneoplastic tumour suppressor activity, such as PCDH10 in various lesions carcinomas [15,16] and PCDH8 in breast cancer [17]. Functional We used combined bisulfite and restriction analysis (COBRA) to data presented here suggest that proteins encoded by the validate and characterise methylation in normal and tumour chromosome 5q31 PCDHs modulate the Wnt pathway and are DNAs. Methylation profiles were consistent with the microarray candidate Wilms’ tumour suppressor genes. data, and locus-wide hypermethylation was observed in all tumours and WiT49 cells (Figure 1C and 1D). We also analysed Results an additional thirteen PCDHs that were identified as hypermethy- lated by MeDIP-chip, demonstrating that hypermethylation varied A large hypermethylated domain at 5q31 in Wilms’ between individual tumours (Figure 1C and 1D). Of 19 PCDHG@ tumour genes, 15 were hypermethylated according to COBRA and array Following methylated DNA immunoprecipitation (MeDIP, analysis, together with 15/16 PCDHB@ genes and 13/15 Figure S1A), the efficacy and specificity of the MeDIP protocol PCDHA@ genes. Bisulfite sequencing data was consistent with was verified using quantitative real-time polymerase chain reaction COBRA and MeDIP-chip data (Figure 1E). In total, we analysed (PCR) of a constitutively methylated CpG island (CGI) at the H19 38 WTs (Table S2) and all display hypermethylation of multiple imprinting control region, demonstrating successful enrichment PCDHs, with many PCDHs showing a very high frequency of relative to input DNA. Specificity for methylated CGIs was shown hypermethylation in tumours. These include PCDHGA3, hyper- by PCR of the RASSF1 59 CGI. Amplification of a non-CGI methylated in 24/27 tumours (89%), PCDHGB4 (23/25, 92%), sequence (an intragenic region of the TBP gene), was used as a PCDHGA2 (11/13, 85%) and PCDHB3 (21/33, 64%). COBRA negative control for methylated DNA enrichment (Figure S1B). analyses of PCDHAC1, PCDHAC2, PCDHB1, PCDHGC3 and Validated MeDIP DNA samples from human fetal kidney and PCDHGC4 showed no evidence of hypermethylation, consistent Wilms’ tumours were then hybridised to Nimblegen HG18 Refseq with MeDIP-chip data (Figure 1B and 1C and Table S2). promoter microarrays (MeDIP-chip) to detect tumour-specific Analysis of PCDH methylation in microdissected perilobar alterations in methylation. nephrogenic rests, presumptive WT precursor lesions, revealed no Microarray data from 5 sporadic WTs were examined to PCDH hypermethylation (Figure S3) but hypermethylation was identify promoters displaying tumour-specific hypermethylation evident at the PCDHGA3 and PCDHGB4 genes in a set of stromal- (P,0.05, t-test). Overall we found 2043 hypermethylated genes. predominant tumours (Figure S4). Taken together our data show Significantly, a large subset of these targets were found in 2 or that hypermethylation occurs at high frequencies in all WT more tumours; pairwise comparisons showed that each individual subtypes, and strongly suggests that de novo methylation arises

PLoS Genetics | www.plosgenetics.org 2 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

Figure 1. A large hypermethylated domain on chromosome 5q31 in Wilms’ tumour. (A) The MeDIP-chip tumour/normal (T/N) signal ratio shown for 2 representative WTs identifies hypermethylation of multiple gene promoters across a domain spanning 800 kb on chromosome 5q31.3 (red box). (B) The MeDIP-chip profile of PCDHA@, PCDHB@ and PCDHG@ in a representative WT. A minority of genes escaping hypermethylation are indicated (black arrowheads). (C) DNA methylation assayed using COBRA. Percentage methylation was analysed by gel densitometry for each gene, and is represented by horizontal bars (black = percentage methylated, white = percentage unmethylated). Data from 6 normal tissues (4 fetal kidneys [FK], postnatal kidney [NK] and fetal brain [FB]) and 9 Wilms’ tumours are shown. Non-PCDH genes flanking the domain and between PCDHB@ and PCDHG@ are also shown. (D) Unsupervised hierarchical clustering of COBRA methylation data from normal and tumour tissues (T), HEK293 and WiT49 cell lines. (E) Bisulfite sequencing of PCDHB3, PCDHB8, SLC25A2, PCDHGA3, PCDHGA6 and PCDHGB4 59 CGIs in normal fetal kidney (FK) and 2 Wilms’ tumours, T43 & T57. doi:10.1371/journal.pgen.1000745.g001 during neoplastic progression from nephrogenic rest to Wilms’ readily detectable in fetal kidney, and consistent suppression of tumour. methylated PCDHG@ transcripts was apparent in tumours relative There was no tumour-specific hypermethylation apparent at to kidney. Of 11 hypermethylated PCDHG@ genes analysed in our upstream and downstream CGIs within 100 kb (eight promoter tumour panel, 9 showed greater than 90% repression, with the CGIs in total) of the PCDH clusters. A non-clustered proto- remaining 2 showing very low basal expression in fetal kidney. For cadherin gene on chromosome 5q31, PCDH1, 300 kb telomeric to many PCDHs, expression in tumours was decreased over 100-fold PCDHG@, was also unmethylated. Two non-PCDH genes are or below the limits of detection (Figure 2A). PCDHGC3 was situated within the methylated domain; SLC25A2 was constitu- consistently unmethylated in WTs (Figure 1C), but PCDHGC3 tively methylated in both normal tissues and tumours and TAF7 expression was lowered in WTs relative to normal kidney was constitutively unmethylated (Figure 1C and Figure S2). Thus, (Figure 2C). Interestingly, WiT49 PCDH methylation and their methylation is unaffected by the surrounding epigenetic expression reflects the general silencing pattern observed with defect, and tumour-specific hypermethylation is specific to WTs (Figure 2B), but in addition to PCDHGC3, PCDHGA6 is also clustered PCDHs. unmethylated in WiT49 cells. Despite the absence of hypermethy- lation, reduced expression of both PCDHGA6 and PCDHGC3 Silencing of PCDH expression in Wilms’ tumour transcripts relative to fetal kidney was apparent in Wit49 cells. Quantitative RT-PCR showed negligible PCDHA@ expression Expression of PCDHGA6 was, however, further suppressed in WTs in kidney. However, PCDHB@ and PCDHG@ transcripts were with PCDHGA6 hypermethylation (Figure 2C).

PLoS Genetics | www.plosgenetics.org 3 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

Figure 2. Silencing of PCDH expression in Wilms’ tumour. (A) Expression levels of genes across the locus in 9 tumours (Texp), relative to the mean of 4 normal fetal kidney samples (N exp). Grey bars are used for genes showing tumour-specific hypermethylation, white bars for unmethylated genes and black bars for the constitutively methylated SLC25A2 gene. (B) Expression levels of 5q31 transcripts correspond with DNA methylation status in WiT49 cells. Expression of unmethylated genes (white bars), constitutively methylated genes (black bar) and hypermethylated PCDHs (grey bars) are shown. (C) Suppression of methylated and unmethylated PCDHs within the chromosome 5q31 LRES. Gene expression levels relative to the house-keeping gene TBP are shown. Horizontal black line, median value; box, interquartile range; whiskers, data range excluding outliers; black dots, outliers (defined as those data points greater than range multiplied by inter-quartile range beyond the box). Grey boxes give samples shown to be hypermethylated, open boxes represent unmethylated samples. doi:10.1371/journal.pgen.1000745.g002

Expression of PCDHB@ genes in tumours was more variable Profiling of histone modifications at active and silenced than the PCDHG@ genes. PCDHB8 and PCDHB15 exhibited genes within the 5q31 LRES strong methylation associated silencing but PCDHB12 was not We compared permissive (histone 3 dimethyl lysine 4, consistently down-regulated in tumours, despite hypermethylation H3K4me2; histone 3 acetyl lysine, H3Ac) and repressive (histone (Figure 2A). All seven methylated PCDHB@ genes analysed were, 3 dimethyl lysine 9, H3K9me2) histone modifications at gene loci however, concordantly silenced in WiT49 cells (Figure 2B). across the PCDH domain (Figure 3) by chromatin immunopre- Unmethylated genes outside the LRES boundary, such as HARS2 cipitation analysis (ChIP). Genes located outside the PCDH and HDAC3, were not suppressed (Figure 2A and 2C). domain were enriched for H3K4me2 and H3Ac. Conversely, To further establish the relationship between methylation and methylated PCDHs were enriched for H3K9me2. Hypermethy- expression, we treated WiT49 cells with 5-azacytidine, which lated PCDHA@, PCDHB@ and PCDHG@ genes all showed induced the expression of epigenetically silenced PCDHB@ and repressive histone modifications. As expected, the SLC25A2 PCDHG@ genes (Figure S5); in contrast, genes located outside the (constitutively methylated) and TAF7 (unmethylated) genes were hypermethylated domain, which showed unaltered expression in associated with repressive and permissive modifications, respec- tumours (Figure 2), were unaffected by 5-azacytidine. This tively. The PCDHB6, PCDHGA6 and PCDHGC3 genes, which substantiates the mechanistic link between methylation levels were not hypermethylated in the WiT49 cell line, all had an active and gene expression at this locus. chromatin profile. In summary, our expression analyses demonstrate that Comparison of gene expression levels, DNA methylation and PCDHB@ and PCDHG@ expression occurs in human kidney, histone marks (Figure 3, right panels) shows that DNA hypermethy- and that epigenetic silencing of gene expression occurs in WT. lation and silencing correlate with diminished H3Ac and

PLoS Genetics | www.plosgenetics.org 4 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

Figure 3. Hypermethylation across the chromosome 5q31 LRES is associated with specific histone modifications. Bar charts (left) show ChIP–quantitative PCR measuring relative levels of specific histone modifications at individual gene loci across the 5q31 locus in WiT49 cells. (A) H3Ac ChIP, (B) H3K4me2 ChIP, (C) H3K9me2 ChIP, all expressed relative to input DNA. Scatter plots (right) show the relationship between relative gene expression levels and histone modifications at each gene. The x-axis shows specific histone levels relative to input DNA, and the y-axis shows mRNA expression in WiT49 cells (Texp) relative to average mRNA expression in 4 fetal kidney samples (Nexp). Grey bars/datapoints signify genes hypermethylated in WiT49, open bars/datapoints show unmethylated genes, and the black bar/datapoint shows constitutively methylated SLC25A2. doi:10.1371/journal.pgen.1000745.g003

H3K4me2, whereas ‘‘active’’ promoters have high H3Ac and similar temporal profiles to Wt1, a mediator of mesenchymal- H3K4me2 levels (Spearman rank order correlation coefficient, epithelial transition (Figure 4A). Postnatally, Pcdh expression r = 0.67, P = 0.003, and r = 0.63, P = 0.006 respectively). H3K9me2 decreases, in contrast to E-cadherin (Cdh1), where high expression shows an opposite pattern, that is high levels at methylated/silenced levels are maintained in the mature organ. As our human genes (Spearman rank order correlation coefficient, r = 20.83, expression analysis was restricted to comparing transcript levels in P = 0.0002). The degree of H3K9me2 enrichment displays fetal kidney and Wilms’ tumours (Figure 2), we also assessed other proportionality to gene silencing (Pearson correlation coefficient, human fetal tissues for PCDHG@ encoded proteins by immuno- r = 0.88). Thus tumour-specific DNA hypermethylation is strongly blotting in order to confirm abundant expression in fetal kidney. linked with specific, repressive chromatin modifications, whereas High expression was also evident in lung and brain, compared unmethylated genes within, and flanking the region maintain an with moderate expression in gut. On longer exposure, low active chromatin configuration. expression was also apparent in liver and spleen (Figure 4B). In order to gain further insight on cell-type specific expression, PCDH expression in renal development and we analysed Pcdhb@ and Pcdhg@ transcript levels during epithelial differentiation differentiation of rat metanephric mesenchymal cells in organ Previous studies have indicated that PCDH expression is largely culture. This system has been shown to accurately reflect early restricted to neuronal tissues [10]. As we found abundant differentiation in embryonic kidney [19]. Freshly isolated meta- expression of PCDHB@ and PCDHG@ genes in human fetal nephric mesenchymal cells were found to express high Pcdh levels, kidney (Figure 2), we examined Pcdhb@ and Pcdhg@ expression and epithelial differentiation induced by growth factors was during murine kidney development. Similar to humans, Pcdha@ accompanied by down-regulation of Pcdhga6, Pcdhga12 and Wt1. genes were expressed predominantly in brain, with negligible Pcdhga2 and Pcdhgc3 were also decreased upon differentiation, levels in kidney. In contrast, Pcdhb@ and Pcdhg@ expression was but to a lesser extent (Figure 4C). As expected, a sharp rise abundant in kidney, with expression comparable to brain and in Cdh1 expression was observed, consistent with increasing exceeding placenta, liver and spleen. Transcript levels followed epithelialisation.

PLoS Genetics | www.plosgenetics.org 5 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

Figure 4. Developmental expression patterns of PCDHs. (A) PCDH transcript levels in mouse developmental tissues. Quantitative real-time expression analysis, relative to Tbp, in placenta (P), and E17.5 mouse fetal liver (Li), spleen (S), brain (B), and kidney (K) (E17.5, postnatal 0.5 week, 1 week, and 3 week) using assays specific for the constant region exons of Pcdha@ and Pcdhg@, and individual Pcdh transcripts. Expression of Wt1 and Cdh1 are also shown (black bars). (B) Immunoblotting of human fetal tissue proteins with pan c-PCDH or actin for loading control. Samples are kidney (K), liver (Li), lung (Lu), spleen (S), brain (B), and gut (G). (C) Gene expression changes accompanying epithelial differentiation of rat metanephric mesenchyme following Lif, Fgf2, and Tgfa treatment. Quantitative real-time expression analysis, relative to Tbp, is shown for freshly dissected rat metanephric mesenchyme (MM) and differentiating mesenchyme (DM). (D) Immunohistochemical analysis of 1-day postnatal murine kidney with towards Wt1 and c-PCDHs. Blastema (b), primitive glomeruli (pg), and ureteric buds (u) are labelled. The control panel shows a section where the primary antibody has been omitted. Bar = 50 mm. doi:10.1371/journal.pgen.1000745.g004

PLoS Genetics | www.plosgenetics.org 6 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

Comparison of expression levels in murine developmental samples and rat mesenchymal cells suggests that the mesenchyme may be the principal cellular component expressing PCDHs in developing kidney. We therefore assessed expression of Pcdhg@ encoded proteins (c-PCDHs) immunohistochemically in postnatal day 1 murine kidney (Figure 4D). High expression was evident in the blastemal cells, with decreasing and more variable expression apparent in tubules and parietal epithelia. Subcellular staining was variable and included nuclear staining, which is consistent with a role for the c-PCDH intracellular domain in gene regulation [20,21]. Expression in the ureteric bud and visceral epithelia was low or absent. Thus c-PCDHs proteins are evident in the murine equivalent of the presumptive multipotent cell of origin for WT. This is consistent with PCDHs having a role in kidney development and Wilms’ tumorigenesis.

Modulation of canonical Wnt signalling by PCDHs Little is known about PCDH cellular functions, but a member of the PCDH superfamily, PCDH-PC, was previously shown to positively regulate b-catenin/TCF signalling [22]. The Wnt signalling pathway, which utilises the b-catenin/TCF transcrip- tional complex to programme developmental gene expression, is essential for nephrogenesis [23]. Constitutive Wnt signalling brought about by compromised b-catenin function is also involved in several cancers including WT [24]. We therefore assessed the possible effects of c-PCDH knockdown on b-catenin/TCF mediated transcription using luciferase reporter plasmids which contain a minimal promoter adjacent to 7 tandem TCF binding sites (Super8xTOPFLASH, [25]). Short-interfering RNAs (siR- NAs) were designed to target the constant region sequences of PCDHG@ and transfected into WiT49 cells. Although WiT49 cells show extensive hypermethylation across the PCDH locus, specific PCDHs (for example, PCDHGA6 and PCDHGC3) escape hyper- methylation and are expressed, albeit at lower levels relative to fetal kidney (Figure 2C); WiT49 cells also exhibit intermediate Wnt signalling activity in the absence of b-catenin mutations, and are therefore suitable for investigating PCDH effects on the Wnt pathway. Figure 5. PCDH effects on Wnt signalling. (A) Enhanced b-catenin/ TCF activity following c-PCDH knockdown induced by PCDHG@ The activity of the b-catenin/TCF reporter was increased by constant region targeting siRNA, measured with Super8xTOPFLASH knockdown of c-PCDHs, suggesting that c-PCDHs negatively reporter (TOP). Super8xFOPFLASH (FOP) reporter is a negative control. influence the canonical Wnt pathway. A second siRNA targeting a RLU, relative luciferase units. Gamma-PCDH and b-catenin knockdowns different sequence within the PCDHG@ constant region gave are verified by immunoblotting (IB) in the inset, which also similar results, negating the possibility of off-target effects (data not demonstrates increased cellular b-catenin accompanying c-PCDH knockdown. (B) Quantitative real-time expression analysis, relative to shown). Reporter upregulation was abolished by CTNNB1 siRNA TBP, showing induction of the Wnt pathway target genes CCND1, CMYC, co-transfection, demonstrating that enhanced TCF-mediated and PAX8 accompanying c-PCDH knockdown; altered expression of activation was b-catenin-dependent (Figure 5A). This was further WT1 is also shown. (C) Repression of b-catenin/TCF reporter activity supported by immunoblot analysis which showed that knockdown accompanying PCDH expression in WiT49, HCT116, and Wnt3a- of c-PCDHs was accompanied by elevated b-catenin protein levels conditioned medium treated HEK293 (HEK293+CM) cells. A plasmid containing a cDNA encoding dominant-negative TCF4 (TCF4-DN) is also (Figure 5A, inset), although CTNNB1 transcript levels were shown as a positive control. Cells were co-transfected with PCDH unchanged (data not shown). Knockdown of c-PCDHs also expression vectors and Super8xTOPFLASH (TOP) or Super8xFOPFLASH induced expression of known b-catenin/TCF target genes CCND1, (FOP) and luciferase activity measured. RLU, relative luciferase units. CMYC and PAX8, as well as repressing transcription of WT1 doi:10.1371/journal.pgen.1000745.g005 (Figure 5B). We also individually over-expressed PCDHGA2, PCDHGA6 and PCDHGA12 in WiT49 cells, human embryonic activity. Taken together, our data implicate PCDHG@ encoded kidney 293 cells (HEK293) treated with Wnt3a conditioned media proteins in negative modulation of canonical Wnt signalling. and HCT116 cells to evaluate effects on b-catenin/TCF reporter As our epigenetic and functional analyses allude to a tumour activity (Figure 5C). The latter cell-line is derived from a colorectal suppressor function for PCDHs, we evaluated the effect of ectopic cancer with an activating b-catenin mutation [26,27] and also PCDH expression on tumour-related phenotype using cell-culture displays PCDH hypermethylation (Figure S6). PCDH-mediated based assays commonly used for assessing tumour suppressor gene suppression of b-catenin/TCF reporter activity was evident in all function, that is inhibition of colony formation and growth in soft cell-lines, especially Wnt3a-treated HEK293 and HCT116, both agar [15]. Transfection of WiT49 cells with PCDHGA2, PCDHGA6 of which have high Wnt signalling. As expected, expression of a or PCDHGA12 resulted in moderate PCDH protein expression dominant-negative form of TCF4 also strongly reduced reporter leading to approximately 30–85% suppression of colony formation

PLoS Genetics | www.plosgenetics.org 7 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

(P,0.05 to P,0.005, t-test). Similarly transfection of HCT116 the first report of LRES in childhood tumours. Our data also suggests cells resulted in approximately 70% suppression of colony roles for PCDHs in normal nephrogenesis, including modulation of formation (P,0.01), and transfection of HEK293 cells led to key regulatory pathways such as canonical Wnt signalling. 60–85% suppression (P,0.005) (Figure 6A). Suppression of colony LRES regions have been identified in several adult cancers formation was not a non-specific side-effect of gene over- including, breast [13], colon [11,12], head, neck and lung expression, as transfection of HEK293 cells with CTNNB1 cDNA [14,28,29]. A recent genome-wide analysis of methylation in encoding a degradation-resistant mutant of b-catenin failed to breast cancers showed that multiple agglomerative epigenetic suppress colony formation (Figure S7). aberrations occur, including regions undergoing hypermethylation We also assessed the effect of PCDH expression on anchorage and hypomethylation. Interestingly, the PCDH locus on chromo- independent growth in soft agar using HCT116 cells (Figure 6B). some 5q31 was one of several hypermethylated domains shown in Colony forming efficiency was markedly reduced by PCDHGA2, breast cancer, together with others such as HOXD@ on PCDHGA6 and PCDHGA12 relative to cells transfected with chromosome 2 and HIST1 on chromosome 6 [30]. Such regions expression vector only (approximately 85%–95% suppression were also identified in our analysis (Table S1). Thus, in addition to relative to the vector control). gene-specific epigenetic lesions, our study shows that some, but not Collectively, therefore, our experiments demonstrate that all, LRES domains are conserved between embryonal and adult PCDHs regulate critical transduction and transcription pathways cancers. Also individual genes within an LRES region can show and have growth regulatory properties consistent with tumour tumour-type specific changes, illustrated by PCDHGC3 which is suppressor activity. frequently hypermethylated in breast cancer but which escapes methylation in WTs. The contribution of LRES to tumour Discussion pathology is not well characterised, but transcriptional suppression of multiple genes across a chromosomal region can be considered By conducting genome-wide promoter methylation analysis, we to be functionally analogous to cytogenetic loss. have identified a large cluster of paralogous PCDH genes on Silencing of individual genes within the LRES domains on chromosome 5q31 which undergo hypermethylation in Wilms’ chromosome 2q and 3p in colorectal cancer appears to be tumours. Transcriptional silencing of PCDHs was prevalent in WTs, dependent on a domain-wide non-permissive chromatin configu- and PCDH hypermethylation constitutes the most frequent epige- ration, rather than the methylation status, as unmethylated genes netic silencing event in WT. This putative WT suppressor domain is within these domains and up to 1000 kb away are also suppressed

Figure 6. Growth inhibition by PCDHs. (A) Suppression of WiT49, HCT116, and HEK293 cell colony formation following ectopic expression of PCDH cDNAs. After selection and staining, plates were photographed and colony counts determined for each transfection. Representative plates (left) and mean colony counts (right) are shown, Verification of PCDH protein expression after transfection, together with tubulin as a loading control is shown below the histograms. (B) Inhibition of anchorage-independent growth of HCT116 cells by PCDHs, cells were plated in triplicate and colonies formed after 10–14 days were photographed and counted within 10 random fields. Representative fields are shown (left) together with colony forming efficiency (CFE), expressed as percentage of colonies .50 mm diameter (right). Cell-based assays were repeated at least twice, and representative data are shown. doi:10.1371/journal.pgen.1000745.g006

PLoS Genetics | www.plosgenetics.org 8 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

[11,12]. By contrast, expression of non-PCDH genes at 5q31 Expression of the Pcdhs peaks in the last week of nephrogenesis; (TAF7 and SLC25A2) was strongly linked to methylation status in thereafter, expression decreases, in contrast to Cdh1, which WTs, and the transcriptional status of unmethylated genes encodes the archetypal epithelial adhesion protein, E-cadherin. flanking the LRES was unchanged in tumours. A comparatively Epithelial differentiation of rat metanephric mesenchyme cells in lesser degree of transcriptional suppression is observed for the organ culture was also accompanied by decreasing levels of unmethylated PCDHGC3 and PCDHGA6 genes in WiT49 cells. Pcdhg@ expression. A recent microarray analysis of gene This suppression occurs despite any significant changes in active/ expression with laser captured kidney components showed repressive histone marks and suggests that PCDHGA6 and expression of Pcdhb15 and Pcdhga12 expression attenuating PCDHGC3 are repressed by a non-epigenetic effect such as altered between the cap mesenchyme and renal vesicle [40]. Together feedback regulation resulting from lowered levels of c-PCDH with our expression analyses, this suggests that the Pcdh expression intracellular domain fragments. Similar to the Notch signalling peak in murine nephrogenesis likely reflects the expansion of paradigm, regulated presenilin dependent-processing of the c- nephrogenic progenitors as kidney development nears completion PCDHs generates C-terminal fragments which can localize to the [41]. Our PCDH expression analyses in human fetal kidney, nucleus and autoregulate the c-PCDHs [20,21]. during murine nephrogenesis and in rat metanehpric mesenchyme In contrast to breast cancer [30], our ChIP data shows a strong suggest that PCDHs may have hitherto uncharacterised roles in link between DNA methylation and H3K9me2 at silenced PCDH renal development. Although Pcdhg@ mutant mice, which undergo genes, as reported for the LRES on chromosome 2q14.2 and 3p22 neurodegeneration and neonatal death in less than 12 hours, did [11,12]. Indeed a correlation between degree of silencing and not show a gross kidney phenotype, kidney defects were not H3K9me2 enrichment was apparent, whereas H3K4me2 and explored in detail [42] (Wang & Sanes, personal communication). H3Ac marking is evident in all active genes and lost in silenced The early postnatal lethality observed with homozygous Pcdhg@ genes. This suggests that H3K9me2 plays a role in establishing mutant mice would also preclude full assessment of effects on and maintaining the silenced state, as previously demonstrated for nephrogenesis, as murine kidney development continues in the CDKN2A [31], and that histone 3 acetylation and H3K4me2 first week following birth. Although we were unable to assess renal marks are removed prior to increases in H3K9me2 and DNA defects in Pcdhg@ null mice, we did examine postnatal kidney from methylation. It has been postulated that a significant proportion of heterozygous Pcdhg@ mutant mice (see Text S1), as heterozygous hypermethylated loci in cancer do not arise by adaptive selection mutations of developmental genes such as Wt1 have been shown to but rather are the result of an ‘instructive’ mechanism, via cis- result in end-stage renal disease [43]. Histological examination of targeting of the trans-acting Polycomb group protein-complexes kidneys from 3 month old heterozygotes showed no evidence of [32], and that these loci are pre-marked in normal (unmethylated) overt kidney malformations (Figure S8). However, it will clearly be tissues by histone H3 trimethyl - lysine27 (H3K27me3). The of great interest to analyse a larger heterozygous cohort together instructive mechanism may explain the non-random de novo with embryonic kidney from homozygous Pcdhg@ mutants in methylation of some genes during tumorigenesis [33]. However, future studies. in the case of the 5q31 LRES, a genomic study of human The canonical Wnt signalling pathway is a prerequisite for embryonic stem cells failed to identify pre-marking of the initiating and maintaining mesenchymal to epithelial transitions hypermethylated PCDHs by Polycomb group proteins or during kidney development, and it is also known that mesenchyme H3K27me3 [34]. Additionally, the instructive mechanism predicts with high b-catenin activity fails to form epithelial structures [44]. methylation would be present in pre-cancerous lesions (e.g. Thus attenuation of Wnt signalling is necessary during nephro- colorectal adenomas [32]) and we have shown this is not the case genesis. Importantly, our functional analysis shows that c-PCDHs for PCDHs, which are unmethylated in nephrogenic rests, pre- repress b-catenin/TCF mediated transcription, with lowered cancerous lesions for WT. Therefore the PCDHs do not appear to PCDH leading to elevated b-catenin protein, high b-catenin/ be pre-marked for de novo methylation in WT, indicating that this TCF reporter activity and induced expression of Wnt target genes. molecular lesion is selected for during tumorigenesis. This is also Conversely, ectopic expression of PCDHs was able to suppress b- supported by tumour-type specific variations in hypermethylation catenin/TCF reporter activity in heterologous cell systems. In such as observed for PCDHGC3, as discussed above. contrast to Wnt target genes, WT1 expression levels were reduced Hypermethylation of PCDHs was not detectable in nephrogenic by PCDH knockdown, demonstrating that Wnt target gene rests, consistent with a previous assessment of RASSF1, DNAJC15/ induction is not reflecting a generalised increase in transcription MCJ and TNFRSF25 gene hypermethylation [35]. This is in and that other regulatory networks are also influenced by cellular contrast to gene-specific hypomethylation of the GLIPR1 gene PCDH levels. The significance of these results is underlined by our observed in WTs, where nephrogenic rests display intermediate findings of epigenetic silencing of PCDHs in WT, as this would be methylation levels relative to fetal kidney and WTs [9]. Therefore predicted to lead to elevated b-catenin/TCF activity. In this although the GLIPR1 hypomethylation observed in WTs might regard, it is notable that enhanced b-catenin signalling in WTs is reflect an expansion of oncofetal cells lacking GLIPR1 methylation, observed more frequently than CTNNB1 and WTX mutations in hypermethylation of PCDHs and other tumour suppressor genes WTs [45,46] and that CTNNB1 mutation is, like PCDH silencing, a appears to represent a later, tumour-specific lesion. Expression of late event in Wilms’ tumorigenesis. Although our analysis of PCDHs in blastemal cells, together with our methylation analysis PCDHs on Wnt signalling can only approximate the permuta- of nephrogenic rests, also negates the possibility that Wilms’ tional silencing in WTs, our results prompt the hypothesis that the tumour PCDH hypermethylation can be attributed to clonal canonical Wnt pathway is modulated by PCDHs, and that in expansion of progenitors with cell-type specific methylation. normal nephrogenesis, elevated PCDHs serve to downregulate b- Genetic lesions in WT known to be late events include catenin activity, thereby permitting completion of epithelial chromosome 16q loss of heterozygosity [36] and CTNNB1 differentiation. Epigenetic silencing of PCDHs might contribute mutations [37]; interestingly, the CTCF gene locates to 16q, is to deregulation of Wnt signalling, and a failure of mesenchymal to mutated in some WTs [38] and the encoded epiregulatory protein epithelial transition resulting in persistence of a progenitor cell has multiple binding sites across the PCDH locus [39], suggesting pool and consequent Wilms’ tumorigenesis. PCDHs may also have that aberrant CTCF function may be involved in LRES. a role in the aetiology of other cancers, such as breast cancer,

PLoS Genetics | www.plosgenetics.org 9 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor where PCDH hypermethylation is prevalent [30] and activation of For PCDH over-expression studies, cDNAs were obtained by Wnt/b-catenin signalling occurs, despite mutations of Wnt RT-PCR from fetal kidney RNA and cloned into pCDNA3.1/Zeo pathway components being rare [47]. (Invitrogen). For b-catenin/TCF reporter assays, 105 cells/well Although the mechanisms by which PCDHs influence pathways were seeded in 24-well plates and transfected with 400 ng of such as Wnt signalling require delineation, we note that, as well as expression constructs together with 100 ng Super 8xTOPFLASH encoding a nuclear moiety capable of regulating gene expression or Super 8xFOPFLASH using Fugene 6 (Roche), and assayed for directly [20,21], a- and c-PCDHs have recently been reported to luciferase as described above. For colony formation assays, negatively regulate proline-rich tyrosine kinase 2 (PYK2) [48] 2.56105 HCT116 and WiT49 cells were transfected with 1 ug which was previously shown to phosphorylate b-catenin [49]. As of expression plasmids in 6-well plates and plated in triplicate in phospho-regulation of b-catenin can promote interactions with 10 cm dishes 48 hours after transfection. Selection was performed transcriptional co-activators [50], we speculate that elevated with 200 mg/ml (HCT116) or 100 mg/ml (WiT49) of zeocin PYK2 activity may arise as a consequence of PCDH silencing (Invitrogen) for 2 weeks after which colonies were methylene blue and thereby lead to a shift of the b-catenin adhesion/signalling stained and counted using ImageJ software (http://rsbweb.nih. balance. This and other downstream consequences of PCDH gov/ij/). Colony formation assays were repeated at least twice. silencing warrant intensive future study. Growth in soft agar was assessed essentially as previously described [15]. Briefly, 2.56104 transfected HCT116 cells were Materials and Methods suspended in DMEM medium containing 10% fetal calf serum, 0.35% agar, and 200 mg/ml zeocin. The suspension was then Ethical statement layered on 6 cm plates containing DMEM medium containing All human tissues were acquired with appropriate local research 10% fetal calf serum, 0.7% agar, and 200 mg/ml zeocin. Plating ethics committee approval and all animal procedures were was carried out in triplicate and repeated at least twice. Cells were conducted in accordance with regulations (UK Home Office) fed every 4–5 days, and after 10–14 days growth, colonies of and international standards on animal welfare. greater than approximately 50 mm within 10 microscopic fields were counted under a phase contrast microscope. Colony forming Patient samples efficiency is presented as percentage of colonies larger than 50 mm All tissues were obtained as snap frozen samples from the Bristol within total cells. Children’s Hospital, the Royal Marsden Hospital and the University of Heidelberg Children’s Hospital. Human fetal tissue Methylated–DNA immunoprecipitation (MeDIP) and samples were from 19 to 31 weeks of gestation. Details of clinical chromatin immunoprecipitation (ChIP) samples are given in Table S3. High molecular weight genomic DNAs were extracted from tissues using standard phenol-chloroform techniques and fragment- Cell culture, transfections, and reporter assays ed to a size range of 200–500 base pairs using a Diagenode WiT49 [18], HCT116 [9] and HEK293 cell-lines (adenovirus Bioruptor. Four micrograms of sonicated genomic DNA and 20 mg transformed human embryonic kidney cells [51]) were cultured anti-5-methyl cytidine monoclonal antibody (Eurogentec, Lie`ge, using standard methods in Dulbecco’s modified Eagle’s medium Belgium) were incubated at 4uC overnight in immunoprecipitation (DMEM) supplemented with 10% fetal calf serum, 2 mmol/l L- buffer, and then for a further 2 hours with goat anti-mouse IgG glutamine, 0.1 mg/ml penicillin/streptomycin, at 37uC under 5% magnetic beads (N.E. Biolabs). After purification, MeDIP DNA was CO2. L/Wnt3a fibroblast cell lines (ATCC, Manassas, VA) were blunt-ended with T4 DNA polymerase (N.E. Biolabs) and ligation- grown in DMEM supplemented with 10% fetal calf serum, and mediated PCR (LM-PCR) was carried out as described [52,53]. Wnt3a conditioned medium was prepared according to the DNAs were then sent to Nimblegen for labelling and hybridization protocol provided by ATCC (http://www.atcc.org). to Nimblegen HG18 Refseq promoter arrays. Data was analysed For knockdown analyses, cells were transfected with Dharma- using ChipMonk v1.2.1 tiling array analysis software (Dr Simon FECT DUO (Dharmacon, Inc.) with 50 nM total of ON- Andrews, Babraham Institute, Cambridge UK, http://www. TARGETplus short interfering RNA (siRNA) duplexes. For b- bioinformatics.bbsrc.ac.uk/projects/chipmonk). To identify hyper- catenin/TCF reporter activity assays, 105 cells/well were seeded in methylated CGIs, a log2ratio cut-off of 1.5 and a window of 500 bp 24-well plates and 100 ng of Super 8xTOPFLASH or Super was used to carry out the replicate t-test (P,0.05) on probes within 8xFOPFLASH reporter plasmids were co-transfected with siRNAs 200 bp of predicted CpG islands (http://genome.ucsc.edu). Addi- and 100 pg of pRL-SV40 to normalise for transfection efficiency. tional statistical analysis was carried out using R (http://www. Super 8xFOPFLASH is a negative control for Super 8xTOP- r-project.org/). The array data reported in this paper have been FLASH containing mutated TCF binding sites [25]. Luciferase deposited in the Gene Expression Omnibus (GEO) database, samples were assayed after 48 hours using Dual-luciferase reporter (http://www.ncbi.nlm.nih.gov/geo) under accession number kit (Promega) and a Modulus Luminometer (Turner Biosystems). GSE15027. Experiments were performed at least twice in triplicate. WiT49 chromatin marks were assessed using a ChIP Kit For rat mesenchymal organ culture, dissected fresh metanephric (Upstate Biotechnology) with antibodies for histone 3 dimethyl mesenchyme from E13.5 rat embryos was cultured on collagen lysine 4 (H3K4me2, Upstate Technology), histone 3 dimethyl coated transwell filters (Corning) in DMEM/F12 media containing lysine 9 (H3K9me2, Abcam), and histone 3 acetyl lysine (H3Ac, 20 mg/ml penicillin/streptomycin, 10 mg/ml insulin, 10 mg/ml Upstate). Quantification using real-time PCR was carried out transferrin, 10 ng/ml selenium (ITS), 40 ng/ml dexamethasone, using the Stratagene MX3005P QPCR System (La Jolla, CA) 100 ng/ml prostaglandin, 4 ng/ml tri-iodo-l-thyronine, 10 ng/ml along with the PlatinumSYBR green qPCR SuperMix-UDG holo-transferrin. Epithelial differentiation was induced over 5 days (Invitrogen). Reactions volumes of 20 ml contained 10 mlof with 50 ng/ml leukemia inhibitory factor (Lif, Chemicon), 20 ng/ Platinum SYBRgreen qPCR SuperMix-UDG (Invitrogen, Paisley, ml transforming growth factor a (Tgfa, R&D Systems) and 50 ng/ UK), 50 nM ROX reference dye, 0.2 mM forward primer, 0.2 mM ml fibroblast growth factor (Fgf2, R&D Systems) [19], after which reverse primer, and 1.5 ml of ChIP DNA template. Primer RNA was extracted with Tri reagent (Sigma). sequences are available in Table S4.

PLoS Genetics | www.plosgenetics.org 10 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

Methylation and expression analysis amplified using ligation-mediated PCR (LM-PCR), labelled and Up to 1 mg DNA was sodium bisulfite converted using the EZ hybridised with promoter microarrays (B) Real-time RT-PCR of DNA Methylation-Gold Kit (Zymo Research, CA). Amplicons for MeDIP-enriched tumour DNAs (upper). Enrichment was validat- combined bisulfite restriction analysis (COBRA) and bisulfite ed using primers specific to a non-CGI sequence within the TBP sequencing were made using the Hot Start Red Taq PCR system gene (grey bars), the constitutively methylated H19 imprinting (Sigma). Primers and restriction enzymes are available in Table control region (black bars), and selective enrichment of the S4. For bisulfite sequencing, PCR products were cloned into the methylated RASSF1 59-CGI in a methylated and unmethylated pGEM-T-easy cloning vector (Promega), and fluorescently tumour (open bars). MeDIP DNA was quantified relative to input sequenced using standard M13 primer sequences. DNA. The lower panel shows COBRA confirmation of H19 For comparative quantitative Real-Time RT-PCR, 1 mgof imprinting control region and the RASSF1 59-CGI methylation DNAse-treated (TURBO DNA-free, Ambion Inc, TX) total RNA status. Arrowheads show methylated (M) and unmethylated (UM) DNA fragments; presence or absence of restriction enzyme is was reverse-transcribed with oligo(dT)20 at 50uCfor1hrusingthe ThermoScript RT-PCR System (Invitrogen). Real-time PCR was indicated (+/2). (C) The tumour (T) versus normal fetal kidney (N) performed using the Stratagene MX3005P QPCR System (La Jolla, signal ratio (y-axis) from 17,777 promoter-associated probes are CA) along with the PlatinumSYBR green qPCR SuperMix-UDG plotted according to their physical location on (x- (Invitrogen). Reactions volumes of 20 ml contained 10 ml of Platinum axis) for 5 WTs. Hypermethylation at 5q31 is indicated by the SYBRgreen qPCR SuperMix-UDG (Invitrogen, Paisley, UK), vertical arrow. 50 nM ROX reference dye, 0.2 mMforwardprimer,0.2mM Found at: doi:10.1371/journal.pgen.1000745.s001 (4.67 MB TIF) reverse primer, and 2.5 ml of 1:10 diluted cDNA template. Primer Figure S2 Methylation analysis of genes neighbouring the sequences are available in Table S4. Thermal cycling consisted of an chromosome 5q31 PCDH cluster. Arrowheads show methylated initial incubation step of 50uC for 2 minutes and a denaturation step (M) and unmethylated (UM) DNA fragments; presence or absence of 95uC for 10 minutes. This was followed by 40 cycles of 95uC/15 of restriction enzyme is indicated (+/2).M+, in vitro methylated seconds, 58uC/30 seconds, 72uC/30 seconds. Gene expression was DNA (A) Distal neighbours of the clustered PCDHs in normal and quantified by comparative Ct method, normalizing values to the tumour tissues. COBRA analysis of CD14, TMCO6, NDUFA2 and housekeeping gene TBP. All assays were performed in duplicate. WDR55 59-CGIs (located -153, -147, -139, and -119 kbp upstream of the PCDH clusters, respectively). FK, 22-week fetal kidney; Immunoblotting and immunohistochemistry WTs, five pooled WT DNAs. (B) 59-CGI methylation analysis of Protein extraction and immunoblotting were carried out the non-clustered PCDH1 gene (located 366 kbp downstream of essentially as previously described [2] with primary antibodies to the PCDH clusters) was carried out on eleven WTs using COBRA. b-catenin (Cell signalling), a-tubulin (Sigma) and the c-PCDH 22-week foetal kidney, FK; FK2, 16-week foetal kidney. constant region (Greg Phillips, Mount Sinai School of Medicine, Found at: doi:10.1371/journal.pgen.1000745.s002 (1.85 MB TIF) USA). This pan c-PCDH antibody is a characterised affinity- Figure S3 Methylation analysis of PCDHB6 in WT precursor purified rabbit polyclonal against a GST-fusion protein containing lesions. (A) COBRA analysis of PCDHB6 in DNA extracted from PCDHG@ the constant cytoplasmic domain encoded by [54,55]. fetal kidney (FK), WTs, and associated perilobar nephrogenic rests This region is highly conserved in human and mouse, with one (NR). T, Wilms’ tumours. Arrowheads show methylated (M) and amino acid difference in 124 amino-acids. Briefly, tissues were unmethylated (UM) DNA fragments; presence or absence of lysed in 300 ml of sample buffer (60 mM Tris pH 6.8, 10% restriction enzyme is indicated (+/2). M+, in vitro methylated glycerol, 2% SDS, 5% mercaptoethanol) and 10 mg protein was DNA. (B) Bisulfite sequencing analysis. Black circles represent loaded per well and electrophoresed on a 10% SDS – methylated CpGs and white circles represent unmethylated CpGs. polyacrylamide gel. After electrophoresis the proteins were Found at: doi:10.1371/journal.pgen.1000745.s003 (1.22 MB TIF) transferred to Immobilon-P (Millipore) with a semidry transfer apparatus. The Immobilon-P was then transferred to 5% non-fat Figure S4 PCDH hypermethylation in stromal-predominant dry milk (Tesco) in PBS (milk block) and blocked for a minimum of Wilms’ tumours. COBRA was carried out for PCDHGA3, 1 hour. Primary antibodies were incubated in milk block overnight PCDHGB4, and HDAC3. Arrowheads show methylated (M) and at 4uC, followed by the secondary antibody at room temperature unmethylated (UM) DNA fragments; presence or absence of for an hour. Protein bands were visualised with ECL Plus reagents restriction enzyme is indicated (+/2). M+, in vitro methylated (Amersham Pharmacia Biotech). DNA. For immunohistochemistry, 5 mm sections of CBA x C57Bl/6 Found at: doi:10.1371/journal.pgen.1000745.s004 (2.02 MB TIF) F2 mouse P0 neonatal kidney were fixed for 16 hrs in 4% Figure S5 Pharmacological demethylation of WiT49 cells. paraformaldehyde. Antigen retrieval of deparaffinised sections was Quantitative real-time RT-PCR of 5q31 transcripts, mock-treated performed by microwaving in 0.8 M urea, pH 6.4, followed by (-) or 5-azacytidine treated (+) cells. Grey bars indicate genes indirect immunoperoxidase staining using the Elite ABC kit associated with hypermethylated CGIs, white bars represent genes (Rabbit IgG; Vectastain) according to the manufacturer’s with CpG islands with no detectable methylation and black bars instructions. Primary antibody dilutions were 1:100 for WT1 are used for SLC25A2. HPRT is an X-chromosome housekeeping (6FH2, Dako), and 1:200 for pan c-PCDH. Sections were control gene. DIAPH1 and HDAC3 are located on chromosome counterstained with haematoxylin. Negative control sections 5q31 outside the LRES. Expression data for 3 PCDHB@ genes omitted the primary antibody. and 4 PCDHG@ genes is shown relative to TBP, together with SLC25A2 and TAF7 genes, which are located within the LRES. Supporting Information Induction of the WT hypermethylated control genes RASSF1 and H19 is also shown. Figure S1 Analysis of aberrant methylation in Wilms’ tumours Found at: doi:10.1371/journal.pgen.1000745.s005 (2.41 MB TIF) using Nimblegen Refseq promoter HG18 tiling arrays. (A) The MeDIP-chip workflow used to analyse genome-wide methylation. Figure S6 Hypermethylation of PCDHGA2, PCDHGA6, Methylated DNA purified from normal and tumour DNAs was PCDHGA12, PCDHB6, PCDHGC3 and PCDHGA7 in HCT116

PLoS Genetics | www.plosgenetics.org 11 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor cells demonstrated using COBRA analysis. Arrowheads show Table S2 COBRA methylation summary. Methylation data methylated (M) and unmethylated (UM) DNA fragments; presence from normal and tumour samples ascertained by combined or absence of restriction enzyme is indicated (+/2). bisulphite restriction analysis (COBRA). U, unmethylated; M, Found at: doi:10.1371/journal.pgen.1000745.s006 (1.25 MB TIF) predominantly methylated; m, partially methylated. L and R, tumours of the left and right kidneys, respectively. Figure S7 Suppression of colony formation is not dependent on Found at: doi:10.1371/journal.pgen.1000745.s010 (0.04 MB b non-specific toxicity of transfected genes. Mutant -catenin (Y33, XLS) tyrosine at amino-acid 33) expression does not suppress colony formation in HEK293 cells. HEK293 cells were transfected with Table S3 Clinical details of nephrogenic rests and Wilms’ CTNNB1 cDNA cloned in the same expression vector tumours. Age at diagnosis: m, age in months. Histology: FH, (pcDNA3.1/Zeo) as PCDH constructs. After selection and staining, favorable histology; UH, unfavorable histology; TR, triphasic; B, plates were photographed and colony counts determined for each blastemal predominant; E; epithelial predominant; S, stromal transfection. Representative plates (above) and mean colony predominant; A, anaplastic; T, teratoid; R, regressive. Outcome: counts (below) are shown. Verification of b-catenin protein A, alive; R, relapsed; D, died. WT1 mutation: Y, yes; N, no. WT1 expression after transfection is shown by immunoblotting below mutation details: G/L, germline. Blank entry, not done. the histograms, together with tubulin to control loading. Found at: doi:10.1371/journal.pgen.1000745.s011 (0.02 MB Found at: doi:10.1371/journal.pgen.1000745.s007 (1.35 MB TIF) XLS) Figure S8 Kidneys of heterozygous Pcdhg@ mutant mice show Table S4 Oligonucleotide primers used in this study. no malformations (see Text S1). Histology of three-month old Found at: doi:10.1371/journal.pgen.1000745.s012 (0.04 MB wild-type (wt, n = 2) and heterozygous Pcdhg@ mutant kidneys XLS) (het, n = 3) was examined on cryosections. Staining of adjacent Text S1 Supporting information methods. sections with cresyl-violet (left column) and nuclear fast red (middle Found at: doi:10.1371/journal.pgen.1000745.s013 (0.03 MB and right columns) was used to highlight the cytoarchitecture of DOC) the specimens. The overall morphology of the heterozygous kidneys appeared normal and showed no malformations. Scale Acknowledgments bars = 500 mm. At higher magnifications, findings were compa- rable in three-month old wild-type and heterozygous littermates The authors wish to thank Vitaliy Androshchuk for excellent support, Iryna Withington for assistance with histology, Greg Phillips for pan c- and displayed normal cytoarchitecture in aged heterozygous mice PCDH antibodies, Herman Yeger for WiT49 cells, Randall Moon for (boxed areas are shown in the right column, scale bars = 100 mm). Super8x reporter plasmids, and Zsombor Melegh and Jamie Davies for Found at: doi:10.1371/journal.pgen.1000745.s008 (1.02 MB TIF) helpful discussions. Table S1 Wilms’ tumour hypermethylated genes identified by MeDIP-chip. Summary table of hypermethylated genes identified Author Contributions by MeDIP-chip in five Wilms’ tumours (P,0.05). Ensembl gene Conceived and designed the experiments: K Malik. Performed the ID, gene symbol, and co-ordinates by chromosome and gene start experiments: AR Dallosso, AL Hancock, M Szemes, K Moorwood, L (HG18 genome build) are given. The frequency of hypermethyla- Chilukamarri, HH Tsai, A Sarkar, C Owen, S Malik, M Frank, KW Brown, K Malik. Analyzed the data: AR Dallosso, AL Hancock, KW tion is given in the final column. Blanks in the Gene name column Brown, K Malik. Contributed reagents/materials/analysis tools: J Barasch, represent Refseq unannotated genes. R Vuononvirta, K Pritchard-Jones, B Royer-Pokora, SB Lee, M Frank, A Found at: doi:10.1371/journal.pgen.1000745.s009 (0.23 MB Ward. Wrote the paper: K Malik. Devised array analysis: K Moorwood, XLS) KW Brown.

References 1. Rivera MN, Haber DA (2005) Wilms’ tumour: connecting tumorigenesis and 10. Morishita H, Yagi T (2007) Protocadherin family: diversity, structure, and organ development in the kidney. Nat Rev Cancer 5: 699–712. function. Curr Opin Cell Biol 19: 584–592. 2. Dallosso AR, Hancock AL, Brown KW, Williams AC, Jackson S, et al. (2003) 11. Frigola J, Song J, Stirzaker C, Hinshelwood RA, Peinado MA, et al. (2006) Genomic Imprinting at the WT1 gene involves a novel coding transcript Epigenetic remodeling in colorectal cancer results in coordinate gene (AWT1) that shows deregulation in Wilms’ tumours. Hum Mol Genet 13: suppression across an entire chromosome band. Nat Genet 38: 540–549. 405–415. 12. Hitchins MP, Lin VA, Buckle A, Cheong K, Halani N, et al. (2007) Epigenetic 3. Malik K, Salpekar A, Hancock A, Moorwood K, Jackson S, et al. (2000) inactivation of a cluster of genes flanking MLH1 in microsatellite-unstable Identification of differential methylation of the WT1 antisense regulatory region colorectal cancer. Cancer Res 67: 9107–9116. and relaxation of imprinting in Wilms’ tumor. Cancer Res 60: 2356–2360. 13. Novak P, Jensen T, Oshiro MM, Wozniak RJ, Nouzova M, et al. (2006) 4. Ogawa O, Eccles MR, Szeto J, McNoe LA, Yun K, et al. (1993) Relaxation of Epigenetic inactivation of the HOXA gene cluster in breast cancer. Cancer Res Insulin-Like Growth Factor-II Gene Imprinting Implicated in Wilms’ Tumour. 66: 10664–10670. Nature 362: 749–751. 14. Rauch T, Wang Z, Zhang X, Zhong X, Wu X, et al. (2007) Homeobox gene 5. Brown KW, Power F, Moore B, Charles AK, Malik KT (2008) Frequency and methylation in lung cancer studied by genome-wide analysis with a microarray- timing of loss of imprinting at 11p13 and 11p15 in Wilms’ tumor development. based methylated CpG island recovery assay. Proc Natl Acad Sci U S A 104: Mol Cancer Res 6: 1114–1123. 5527–5532. 6. Zhang L, Anglesio MS, O’Sullivan M, Zhang F, Yang G, et al. (2007) The E3 15. Ying J, Li H, Seng TJ, Langford C, Srivastava G, et al. (2006) Functional ligase HACE1 is a critical chromosome 6q21 tumor suppressor involved in epigenetics identifies a protocadherin PCDH10 as a candidate tumor suppressor multiple cancers. Nat Med 13: 1060–1069. for nasopharyngeal, esophageal and multiple other carcinomas with frequent 7. Wagner KJ, Cooper WN, Grundy RG, Caldwell G, Jones C, et al. (2002) methylation. Oncogene 25: 1070–1080. Frequent RASSF1A tumour suppressor gene promoter methylation in Wilms’ 16. Yu J, Cheng YY, Tao Q, Cheung KF, Lam CN, et al. (2009) Methylation of tumour and colorectal cancer. Oncogene 21: 7277–7282. protocadherin 10, a novel tumor suppressor, is associated with poor prognosis in 8. Morris MR, Hesson LB, Wagner KJ, Morgan NV, Astuti D, et al. (2003) patients with gastric cancer. Gastroenterology 136: 640–651. Multigene methylation analysis of Wilms’ tumour and adult renal cell 17. Yu JS, Koujak S, Nagase S, Li CM, Su T, et al. (2008) PCDH8, the human carcinoma. Oncogene 22: 6794–6801. homolog of PAPC, is a candidate tumor suppressor of breast cancer. Oncogene 9. Chilukamarri L, Hancock AL, Malik S, Zabkiewicz J, Baker JA, et al. (2007) 27: 4657–4665. Hypomethylation and aberrant expression of the glioma pathogenesis-related 1 18. Alami J, Williams BR, Yeger H (2003) Derivation and characterization of a gene in Wilms tumors. Neoplasia 9: 970–978. Wilms’ tumour cell line, WiT 49. Int J Cancer 107: 365–374.

PLoS Genetics | www.plosgenetics.org 12 November 2009 | Volume 5 | Issue 11 | e1000745 PCDH Domain Silencing in Wilms’ Tumor

19. Schmidt-Ott KM, Masckauchan TNH, Chen X, Hirsh BJ, Sarkar A, et al. 38. Filippova GN, Qi CF, Ulmer JE, Moore JM, Ward MD, et al. (2002) Tumor- (2007) beta-catenin/TCF/Lef controls a differentiation-associated transcription- associated zinc finger mutations in the CTCF transcription factor selectively al program in renal epithelial progenitors. Development 134: 3177–3190. alter its DNA-binding specificity. Cancer Res 62: 48–52. 20. Haas IG, Frank M, Veron N, Kemler R (2005) Presenilin-dependent processing 39. Kim TH, Abdullaev ZK, Smith AD, Ching KA, Loukinov DI, et al. (2007) and nuclear function of gamma-protocadherins. J Biol Chem 280: 9313–9319. Analysis of the vertebrate insulator protein CTCF-binding sites in the human 21. Hambsch B, Grinevich V, Seeburg PH, Schwarz MK (2005) Gamma- genome. Cell 128: 1231–1245. Protocadherins, presenilin-mediated release of C-terminal fragment promotes 40. Brunskill EW, Aronow BJ, Georgas K, Rumballe B, Valerius MT, et al. (2008) locus expression. J Biol Chem 280: 15888–15897. Atlas of gene expression in the developing kidney at microanatomic resolution. 22. Yang XZ, Chen MW, Terry S, Vacherot F, Chopin DK, et al. (2005) A human- Dev Cell 15: 781–791. and male-specific protocadherin that acts through the Wnt signaling pathway to 41. Kobayashi A, Valerius MT, Mugford JW, Carroll TJ, Self M, et al. (2008) Six2 induce neuroendocrine transdifferentiation of prostate cancer cells. Cancer Res defines and regulates a multipotent self-renewing nephron progenitor population 65: 5263–5271. throughout mammalian kidney development. Cell Stem Cell 3: 169–181. 23. Schmidt-Ott KM, Barasch J (2008) WNT/beta-catenin signaling in nephron 42. Wang XZ, Weiner JA, Levi S, Craig AM, Bradley A, et al. (2002) Gamma progenitors and their epithelial progeny. Kidney Int 74: 1004–1008. protocadherins are required for survival of spinal interneurons. Neuron 36: 24. Giles RH, van Es JH, Clevers H (2003) Caught up in a Wnt storm: Wnt 843–854. signaling in cancer. Biochim Biophys Acta 1653: 1–24. 43. Menke AL, Ijpenberg A, Fleming S, Ross A, Medine CN, et al. (2003) The wt1- 25. Veeman MT, Slusarski DC, Kaykas A, Louie SH, Moon RT (2003) Zebrafish heterozygous mouse; a model to study the development of glomerular sclerosis. prickle, a modulator of noncanonical Wnt/Fz signaling, regulates gastrulation J Pathol 200: 667–674. movements. Curr Biol 13: 680–685. 44. Park JS, Valerius MT, McMahon AP (2007) Wnt/beta-catenin signaling 26. Brattain MG, Fine WD, Khaled FM, Thompson J, Brattain DE (1981) regulates nephron induction during mouse kidney development. Development Heterogeneity of malignant cells from a human colonic carcinoma. Cancer Res 134: 2533–2539. 41: 1751–1756. 45. Ruteshouser EC, Robinson SM, Huff V (2008) Wilms tumor genetics: mutations 27. Ilyas M, Tomlinson IP, Rowan A, Pignatelli M, Bodmer WF (1997) Beta-catenin in WT1, WTX, and CTNNB1 account for only about one-third of tumors. mutations in cell lines established from human colorectal cancers. Proc Natl Genes Cancer 47: 461–470. Acad Sci U S A 94: 10330–10334. 46. Koesters R, Ridder R, Kopp-Schneider A, Betts D, Adams V, et al. (1999) 28. Smith LT, Lin M, Brena RM, Lang JC, Schuller DE, et al. (2006) Epigenetic Mutational activation of the beta-catenin proto-oncogene is a common event in regulation of the tumor suppressor gene TCF21 on 6q23-q24 in lung and head the development of Wilms’ tumors. Cancer Res 59: 3880–3882. and neck cancer. Proc Natl Acad Sci U S A 103: 982–987. 47. Howe LR, Brown AM (2004) Wnt signaling and breast cancer. Cancer Biol 29. Liu Z, Zhao J, Chen XF, Li W, Liu R, et al. (2008) CpG island methylator Ther 3: 36–41. phenotype involving tumor suppressor genes located on chromosome 3p in non- 48. Chen J, Lu Y, Meng S, Han MH, Lin C, et al. (2009) Alpha- and Gamma- small cell lung cancer. Lung Cancer 62: 15–22. Protocadherins negatively regulate PYK2. J Biol Chem 284: 2880–2890. 30. Novak P, Jensen T, Oshiro MM, Watts GS, Kim CJ, et al. (2008) Agglomerative 49. van Buul JD, Anthony EC, Fernandez-Borja M, Burridge K, Hordijk PL (2005) epigenetic aberrations are a common event in human breast cancer. Cancer Res Proline-rich tyrosine kinase 2 (Pyk2) mediates vascular endothelial-cadherin- 68: 8616–8625. based cell-cell adhesion by regulating beta-catenin tyrosine phosphorylation. 31. Bachman KE, Park BH, Rhee I, Rajagopalan H, Herman JG, et al. (2003) J Biol Chem 280: 21129–21136. Histone modifications and silencing prior to DNA methylation of a tumor 50. Daugherty RL, Gottardi CJ (2007) Phospho-regulation of Beta-catenin adhesion suppressor gene. Cancer Cell 3: 89–95. and signaling functions. Physiology 22: 303–309. 32. Keshet I, Schlesinger Y, Farkash S, Rand E, Hecht M, et al. (2006) Evidence for 51. Graham FL, Smiley J, Russell WC, Nairn R (1977) Characteristics of a human an instructive mechanism of de novo methylation in cancer cells. Nat Genet 38: cell line transformed by DNA from human adenovirus type 5. J Gen Virol 36: 149–153. 59–74. 33. Schlesinger Y, Straussman R, Keshet I, Farkash S, Hecht M, et al. (2007) 52. Weber M, Davies JJ, Wittig D, Oakeley EJ, Haase M, et al. (2005) Polycomb-mediated methylation on Lys27 of histone H3 pre-marks genes for de Chromosome-wide and promoter-specific analyses identify sites of differential novo methylation in cancer. Nat Genet 39: 232–236. DNA methylation in normal and transformed human cells. Nat Genet 37: 34. Lee, TI, et al. (2006) Control of developmental regulators by Polycomb in 853–862. human embryonic stem cells, Cell 125: 301–13. 53. Squazzo SL, O’Geen H, Komashko VM, Krig SR, Jin VX, et al. (2006) Suz12 35. Ehrlich M, Jiang G, Fiala E, Dome JS, Yu MC, et al. (2002) Hypomethylation binds to silenced regions of the genome in a cell-type-specific manner. Genome and hypermethylation of DNA in Wilms tumors. Oncogene 21: 6694–6702. Res 16: 890–900. 36. Charles AK, Brown KW, Berry PJ (1998) Microdissecting the genetic events in 54. Phillips GR, Tanaka H, Frank M, Elste A, Fidler L, et al. (2003) Gamma- nephrogenic rests and Wilms’ tumor development. Am J Path 153: 991–1000. protocadherins are targeted to subsets of synapses and intracellular organelles in 37. Fukuzawa R, Heathcott RW, More HE, Reeve AE (2007) Sequential WT1 and neurons. J Neurosci 23: 5096–5104. CTNNB1 mutations and alterations of beta-catenin localisation in intralobar 55. Frank M, Ebert M, Shan WS, Phillips GR, Arndt K, et al. (2005) Differential nephrogenic rests and associated Wilms tumours: two case studies. J Clin Pathol expression of individual gamma-protocadherins during mouse brain develop- 60: 1013–1016. ment. Mol Cell Neurosci 29: 603–616.

PLoS Genetics | www.plosgenetics.org 13 November 2009 | Volume 5 | Issue 11 | e1000745